A method for optimizing logistics distribution network

Abstract

The present invention discloses a method for optimizing a logistics distribution network, comprising the following steps: S1, determining a set of candidate points of a distribution center in a city; S2, constructing a logistics distribution network optimization model according to the distribution center; S3, iteratively optimizing the logistics distribution network optimization model using an improved genetic algorithm; S4, using the iteratively optimized logistics distribution network optimization model to layout the logistics distribution network, completing the optimization of the logistics distribution network, and providing an effective reference for the optimization of the logistics distribution network of a third-party logistics enterprise.

Description

Technical Field

[0001] The present invention relates to the field of distribution technology, and in particular to a method for optimizing a logistics distribution network. Background Art

[0002] With the increasing demand of urban residents for transportation services of various commodities, we are faced with the challenges of high land prices and limited land for logistics in cities, which poses an obstacle to the further development of the logistics and distribution industry. At the same time, there are still many underutilized land resources around highway service areas, and relevant management departments are actively seeking new service models to promote the development and utilization of land resources. Our logistics and warehousing facilities are located based on highway construction, which helps to avoid heavy trucks from entering the city and alleviate the problem of urban traffic congestion. In addition, relying on the developed highway network for transportation improves the efficiency of highway use, while also enhancing the timeliness of cargo delivery, achieving a win-win situation of corporate economic benefits and social benefits.

[0003] At present, the research on the optimization of logistics distribution network in highway service areas has a certain foundation, but there are still some problems to be solved:

[0004] 1. Most studies only consider the construction of distribution centers in service areas and regard them as a new node in the existing highway logistics network. The distribution mode has not changed much and the potential of highways as a closed transportation system has not been fully utilized.

[0005] 2. In terms of mathematical model construction, the focus is mainly on the location model of the service area distribution center, while there is relatively little research on subsequent route arrangement;

[0006] 3. The consideration of the total system cost in the objective function is not detailed enough, and the characteristics of highway transportation, such as its strong timeliness and the additional tolls and driving safety factors generated at the same time, are not fully reflected in the constructed model;

[0007] 4. When considering urban distribution, the determination of the distribution and quantity of demand points within the city is usually relatively vague, and simple estimates of direct settings or single-dimensional indicators (such as the permanent population of a district or county) are often used. Summary of the invention

[0008] In view of the above-mentioned deficiencies in the prior art, a logistics distribution network optimization method provided by the present invention solves the problems of the existing model that the potential of the highway as a closed transportation system is not fully utilized, the path arrangement is not detailed, and the total system cost is not considered in detail.

[0009] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: to provide a logistics distribution network optimization method, comprising the following steps:

[0010] S1. Determine the set of candidate points for the city's distribution center;

[0011] S2. Construct a logistics distribution network optimization model based on the distribution center;

[0012] S3, use improved genetic algorithm to iteratively optimize the logistics distribution network optimization model;

[0013] S4. Use the iteratively optimized logistics distribution network optimization model to layout the logistics distribution network and complete the logistics distribution network optimization.

[0014] Further: the step S1 comprises the following sub-steps:

[0015] S11, obtaining data of urban communities and determining original community demand points;

[0016] S12, clustering the original community demand points using the K-medoids clustering algorithm to obtain a new demand point group;

[0017] S13, clustering the new demand point group using the K-medoids clustering algorithm to obtain a number of cluster labels and cluster points;

[0018] S14, select a cluster label, calculate the distance between all the points in the cluster label and each service area, select two areas with the smallest average distance to the points in the cluster as candidate points for serving the demand points in the cluster, and add them to the candidate list;

[0019] S15, repeat step S14 until all cluster labels are traversed;

[0020] S16. De-duplicate the candidate list to obtain a final set of candidate points, which is used as the set of candidate points for the city's distribution center.

[0021] The beneficial effect of the above further scheme is: using the K-medoids clustering algorithm for clustering is very sensitive to outliers in the data compared to the K-means algorithm. Some prominent outliers may seriously distort the clustering results. The characteristic of residential communities is that some communities are far away from the clustering area, thus presenting an "outlier" effect. K-medoids will take the object at the center of the cluster as the center, instead of using the average value of objects in the class as the reference point like K-means, but K-medoids also retains the principle of K-means clustering based on minimizing the sum of the dissimilarity between each object and the class center point, which makes the K-medoids algorithm very suitable for the data set of this article.

[0022] Furthermore, the basic conditions for constructing the logistics distribution network optimization model include:

[0023] The quantity and location of the distribution centers are determined from a set of candidate points;

[0024] The location and demand of each customer demand point served by the distribution center are known quantities;

[0025] The delivery vehicle departs from the distribution center to the customer demand point for delivery, and returns to the distribution center after serving all customers;

[0026] Each customer demand point can only be delivered by one distribution center;

[0027] There is no transfer of goods between distribution centers;

[0028] The delivery vehicles travel at a constant speed on both highway sections and urban roads;

[0029] Customers have time requirements for logistics delivery services, and exceeding the given time frame will incur time penalty costs;

[0030] Each delivery vehicle corresponds to a delivery route, and the cargo load does not exceed the load limit.

[0031] Further: The formula of the logistics distribution network optimization model is:

[0032] MinZ1=min(F1+F2+F3+F4)

[0033]

[0034] Where Z1 represents the total system cost of the logistics distribution network, F1, F2, F3 and F4 represent the leasing and operating cost of the logistics distribution center, the vehicle transportation cost, the customer delivery time penalty cost and the cargo damage cost respectively; S = {0,1,2,…,s} represents the set of distribution centers; C = {0,1,2,…,c} represents the set of ordinary goods customers; D = {0,1,2,…,d} represents the set of cold chain fresh food customers; M = {0,1,2,…,m} represents the set of ordinary goods distribution vehicles in the service area distribution center; N = {0,1,2,…,n} represents the set of cold chain distribution vehicles in the service area distribution center; l ij represents the delivery distance from point i to point j; u j represents the demand for ordinary goods of the jth customer; e j represents the cold chain goods demand of the jth customer; FS i represents the daily rental of the distribution center in the i-th service area; H1 represents the handling fee per unit weight of ordinary goods in the distribution center; H2 represents the handling fee per unit weight of cold chain goods in the distribution center; FC k represents the fixed cost of dispatching the kth ordinary truck; FD krepresents the fixed cost of dispatching the kth cold chain truck; f1 represents the energy consumption coefficient per unit distance of ordinary trucks; f2 represents the energy consumption coefficient per unit distance of cold chain trucks; f3 represents the penalty rate for providing logistics services before the customer's earliest time window; f4 represents the penalty rate for providing logistics services after the customer's latest time window; FRC ij It represents the highway toll required for ordinary vehicles to travel from point i to point j; FRD ij represents the highway toll required for cold chain vehicles to travel from point i to point j; X ijk , Y ij and Z i are all decision variables; [ET j ,LT j ] represents the time window for the jth customer to receive service; T jk represents the time when customer j receives the service from car k; rc j Indicates the damage rate of fresh products; rn j Indicates the damage rate of ordinary goods; p c represents the unit damage cost of cold chain fresh goods; p n Represents the unit damage cost of ordinary goods. Further: The constraints of the logistics distribution network optimization model are:

[0035]

[0036]

[0037] Among them, P k represents the maximum load of the kth ordinary delivery vehicle; Q k represents the maximum load of the kth cold chain delivery vehicle; r ijk represents the load of the kth vehicle when it leaves point i and arrives at point j; r jik represents the load of the kth vehicle when it leaves point j and arrives at point i; T ik represents the time when customer i receives the service of car k, t ik represents the length of time that customer i receives k-vehicle logistics service; t ijk represents the time required for the kth car to get from point i to point j; S j Indicates customer satisfaction, X jik is the decision variable, [ET i ,LT i ] represents the service time window of the distribution center in the i-th service area; T represents the minimum average comprehensive customer satisfaction.

[0038] The beneficial effect of the above further solution is: establishing a two-layer LRP model with the goal of minimizing the total system cost and optimizing the entire logistics distribution network.

[0039] Further: the step S3 comprises the following sub-steps:

[0040] S31, encoding the chromosome according to the number of vehicles, center numbers and customer point numbers in the logistics distribution network optimization model;

[0041] S32, initializing the population according to the encoded chromosome;

[0042] S33, calculating the comprehensive fitness value of chromosomes in the population;

[0043] S34, randomly performing genetic update on the initialized population by selection, crossover or mutation;

[0044] S35, using a local search algorithm to optimize the inherited population to obtain a new population;

[0045] S36. Repeat steps S33-S35 until a preset number of iterations is reached, thereby completing the iterative optimization of the logistics distribution network optimization model.

[0046] Further: In the step S31, the chromosome is encoded by using a non-fixed-length real number encoding method, and the encoded chromosome

[0047] Among them, a is the number of vehicles in use, a=1,2,3,...,m, m is the number of chromosomes; substring C k a={l w0 ,l w1 ,l w2 ,...,l wd}, l w0 is the distribution center number, l w1 ,l w2 ,...,l wd Number the customer's order.

[0048] Furthermore: in the step S34, the genetic algorithm used is a roulette wheel method, which performs roulette wheel division according to the comprehensive fitness values ​​of chromosomes in the population.

[0049] Further: In step S34, the method of using the crossover genetic algorithm to genetically update chromosomes in the population includes the following sub-steps:

[0050] S3401, select two chromosomes in the population as parent chromosomes;

[0051] S3402, randomly select a coding position representing a distribution center in each parent chromosome, and extract the entire substring where it is located as the crossover position;

[0052] S3403, exchange and insert the crossover site into another parent chromosome to form a new chromosome;

[0053] S3404. De-duplicate the new chromosomes to obtain offspring chromosomes, thus completing the genetic update.

[0054] Further: In step S34, the method of genetically updating chromosomes in the population using a variant genetic algorithm includes the following sub-steps:

[0055] S3411, setting the mutation probability, and determining the chromosome to be mutated according to the mutation probability;

[0056] S3412, randomly selecting a number of substrings in the chromosome to be mutated;

[0057] S3413, performing distribution center code bit mutation on each substring;

[0058] S3414. Perform path encoding bit mutation on each substring to obtain offspring chromosomes and complete genetic update.

[0059] The beneficial effects of the present invention are:

[0060] 1. By improving the existing logistics network, multiple beneficial effects are achieved, making the logistics system more efficient, economical and environmentally friendly;

[0061] 2. It can quickly respond to the logistics needs of urban residents for specific goods, improve distribution efficiency, and significantly reduce the total cost of the system;

[0062] 3. This method introduces comprehensive customer satisfaction, which not only improves delivery efficiency but also takes into account customer experience;

[0063] 4. Make full use of the potential of highways as a closed transportation system so that large trucks no longer need to enter urban areas, thereby effectively reducing traffic congestion in cities and improving traffic efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 This is a flow chart of a logistics distribution network optimization method proposed by the present invention.

[0065] Figure 2 This is a conceptual diagram of the logistics distribution network proposed in this invention.

[0066] Figure 3 This is a schematic diagram of the logistics distribution network proposed by the present invention.

[0067] Figure 4 This is the customer time satisfaction function graph.

[0068] Figure 5Schematic diagram of the crossover genetic algorithm process.

[0069] Figure 6 Schematic diagram of the mutation genetic algorithm process.

[0070] Figure 7 This is the optimal delivery route map for ordinary vehicles in the embodiment of the present invention.

[0071] Figure 8 This is the optimal delivery route map for cold chain vehicles in an embodiment of the present invention. DETAILED DESCRIPTION

[0072] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0073] like Figure 1 As shown, in one embodiment of the present invention, a method for optimizing a logistics distribution network is provided, which includes the following steps:

[0074] S1. Determine the set of candidate points for the city's distribution center;

[0075] S2. Construct a logistics distribution network optimization model based on the distribution center;

[0076] S3, use improved genetic algorithm to iteratively optimize the logistics distribution network optimization model;

[0077] S4. Use the iteratively optimized logistics distribution network optimization model to layout the logistics distribution network and complete the logistics distribution network optimization.

[0078] The concept diagram of the logistics distribution network of the present invention is as follows Figure 2 shown.

[0079] In one embodiment of the present invention, step S1 includes the following sub-steps:

[0080] S11, obtaining data of urban communities and determining original community demand points;

[0081] S12, clustering the original community demand points using the K-medoids clustering algorithm to obtain a new demand point group;

[0082] S13, clustering the new demand point group using the K-medoids clustering algorithm to obtain a number of cluster labels and cluster points;

[0083] S14, select a cluster label, calculate the distance between all the points in the cluster label and each service area, select two areas with the smallest average distance to the points in the cluster as candidate points for serving the demand points in the cluster, and add them to the candidate list;

[0084] The service area is a highway service area, which is obtained by crawling keywords using the Amap API in this embodiment.

[0085] S15, repeat step S14 until all cluster labels are traversed;

[0086] S16. De-duplicate the candidate list to obtain a final set of candidate points, which is used as the set of candidate points for the city's distribution center.

[0087] In this embodiment, two layers of K-medoids clustering are included. The first layer of clustering: the first layer of clustering is performed on the original residential community demand points to obtain new demand points.

[0088] This method uses the K-medoids clustering algorithm to cluster the original resident demand points. The K-medoids algorithm is essentially an improvement on the K-means algorithm. The K-means algorithm is an efficient distance-based clustering method, but it is very sensitive to outliers in the data. Some prominent outliers may seriously distort the clustering results. The characteristic of residential communities is that some communities are far away from the clustering area, thus presenting an "outlier" effect. In order to eliminate the impact of some communities being too remote, this method uses the K-medoids algorithm for clustering. K-medoids will use the object at the center of the cluster as the center, instead of using the average value of objects in the class as the reference point like K-means, but K-medoids also retains the principle of K-means clustering based on minimizing the sum of the dissimilarities between each object and the class center point, which makes the K-medoids algorithm very suitable for the data set of this article.

[0089] Second-level clustering: obtain the classification of several cluster centers and data points. To determine the distribution center that serves the demand points in the cluster, calculate the distance between all points in the cluster and each service area. Select the two areas with the smallest average distance to all points in the cluster as candidate points to serve the demand points in the cluster and add them to the candidate list. Finally, deduplicate the service areas in the list to obtain a set of candidate points for the city's distribution center.

[0090] In one embodiment of the present invention, the principle of the K-medoids clustering algorithm is as follows:

[0091] Step 1: Randomly select k points from the total n sample points as medoids;

[0092] Step 2: According to the principle of being closest to the medoids, the remaining nk points are assigned to the class represented by the current best medoids;

[0093] Step 3: For all other points in the i-th class except the corresponding medoids point, calculate the sum of the distances to other points when they are new medoids (Haversine distance) in order, traverse all possibilities, and select the point with the smallest sum of distances as the new medoids;

[0094] Step 4: Repeat the process of Step 2 and 3 until all medoids points no longer change or the maximum number of iterations has been reached;

[0095] Step 5: Get the final k classes.

[0096] In addition, since the clustered data points in this method are the longitude and latitude information of the community, it is different from the Euclidean distance calculation formula between two points in the plane rectangular coordinate system, and the sphere characteristics need to be considered. Therefore, this method uses the classic Haversine formula to calculate the distance between longitude and latitude. The Haversine formula uses a sine function, which can ensure accuracy even if the distance is very small. The specific calculation formula is as follows:

[0097]

[0098] Where d is the distance between any two points on the earth; R is the radius of the earth, which can be taken as an average value of 6371km; φ i ,φ j represents the latitude of two points; Δλ represents the difference in longitude between two points.

[0099] For any angle θ:

[0100] haversin(θ)=sin 2 (θ / 2)=(1-cos(θ)) / 2 (1-2)

[0101] From formula (1-1) and formula (1-2), we can get:

[0102]

[0103] Formula (1-3) is the final distance calculation formula.

[0104] In one embodiment of the present invention, a logistics distribution network based on the expansion of the logistics function of highway service areas (HSA-LDN) is proposed, that is, all distribution centers in the network are established in highway service areas, and transportation is carried out through a developed highway network.

[0105] like Figure 3As shown, the city terminal distribution link is selected for modeling, in which the goods of all distribution centers have been transferred from the previous distribution center to the current distribution center, and now these goods need to be distributed to the next level.

[0106] The location-route problem of the logistics distribution network in this method can be described as follows: After a two-layer clustering algorithm, the urban resident community data is used to determine a number of urban resident demand points, demand quantities, and a number of distribution center candidate point sets. Now it is necessary to select a certain number of ordinary / cold chain distribution centers from the distribution center candidate point set to build distribution services for the urban resident demand points, including distribution and cold chain fresh food distribution.

[0107] In one embodiment of the present invention, the basic conditions for constructing the logistics distribution network optimization model include:

[0108] The quantity and location of the distribution centers are determined from a set of candidate points;

[0109] The location and demand of each customer demand point served by the distribution center are known quantities;

[0110] The delivery vehicle departs from the distribution center to the customer demand point for delivery, and returns to the distribution center after serving all customers;

[0111] Each customer demand point can only be delivered by one distribution center;

[0112] There is no transfer of goods between distribution centers;

[0113] The delivery vehicles travel at a constant speed on both highway sections and urban roads;

[0114] Customers have time requirements for logistics delivery services, and exceeding the given time frame will incur time penalty costs;

[0115] Each delivery vehicle corresponds to a delivery route, and the cargo load does not exceed the load limit.

[0116] In one embodiment of the present invention, the logistics distribution network optimization model is constructed from the perspective of third-party logistics enterprises, taking into account the various costs and comprehensive customer satisfaction required to build the entire urban logistics distribution network based on the logistics function expansion of highway service areas, and establishing a two-layer LRP model with the lowest total system cost as the goal to optimize the entire logistics distribution network. The decision content includes the number and location of distribution centers, the service relationship between distribution centers and various demand points, the type and number of vehicles used from distribution centers to various demand points, and the specific route planning arrangements;

[0117] Among them, the costs of third-party logistics distribution companies include:

[0118] (1) Rental costs

[0119] For third-party logistics companies, this part of the cost is reflected as leasing cost. Different from leasing a general distribution center, since highway service areas have ready-made idle land resources, their land cost will be much lower than building a new distribution center in the city. Therefore, the leasing cost of third-party logistics companies will also be lower, which will be fully considered in the subsequent parameter setting process.

[0120] (2) Operating costs

[0121] In terms of operating costs, the operating costs of a distribution center are similar to those of a general distribution center, mainly including cargo handling costs, water and electricity costs, labor costs, etc.

[0122] (3) Transportation costs

[0123] In addition to the fixed costs generated by the dispatch of different vehicles, the variable costs of the distribution center transportation cost mainly include the fuel consumption of vehicles and highway tolls. Compared with ordinary road transportation, highway transportation maintains high speed, the road surface is smoother, the fuel consumption coefficient of vehicles is lower, and fuel costs are saved. However, the tolls generated are higher, which is an additional expense cost.

[0124] (4) Delivery time penalty cost

[0125] The penalty cost is the penalty imposed for service time outside the customer's preferred time window to keep customer satisfaction at a high level.

[0126] (5) Cargo damage costs

[0127] Any goods may be damaged during transportation. Generally speaking, the damage of goods during the distribution process mainly comes from two aspects. One is that improper loading and unloading leads to squeezing and puncture of goods; the other is damage caused by bumps and collisions during transportation. In addition, the quality of fresh products will decrease with time and temperature, and its characteristics of corruption and deterioration need to be considered. According to the characteristics of different goods, this method customizes different cargo damage rate functions to characterize the loss of goods during transportation. At the same time, because the road surface of the highway is smoother than that of ordinary roads, the road conditions are simpler, and the driving process is smoother and safer. Therefore, the cargo damage rate on the highway should be lower than that on ordinary roads. This method also takes this situation into consideration and designs different cargo damage rates for high-speed and non-high-speed conditions.

[0128] The freshness of fresh products is mainly affected by factors such as temperature, time, and squeezing. This paper introduces the quality spoilage function D(t) = D0e -θt , where D(t) is the freshness of the product at time t, and θ is the spoilage rate of the product.

[0129] Fresh product damage rate rc j It can be expressed as:

[0130]

[0131] Where θ1 is the spoilage rate of fresh products.

[0132] Then the cost of cargo damage is:

[0133]

[0134] For ordinary goods, there is no quality corruption problem, so the damage rate formula for fresh products is not applicable. Here we only consider the linear relationship between the integrity of ordinary goods and the transportation time. As the transportation time increases, the integrity of the package continues to decline. The damage rate formula is as follows:

[0135] rn j =θ2(T jk -T ik ) (2-3)

[0136] Where θ2 is the damage rate of ordinary goods.

[0137] Then the damage cost of ordinary goods is:

[0138]

[0139] Therefore, the total cargo damage cost is:

[0140]

[0141] (6) Overall customer satisfaction

[0142] When considering customer satisfaction, we also take into account customer requirements for delivery timeliness and product quality, and propose a comprehensive customer satisfaction to characterize these two aspects, including customer timeliness satisfaction and product integrity satisfaction.

[0143] Customer satisfaction rate:

[0144] Whether it is ordinary goods or cold chain fresh goods, customers have certain requirements for their delivery time. In order to improve their service quality, logistics companies will pay great attention to customer satisfaction with delivery time. Therefore, this method introduces a linear continuous time satisfaction function to quantify customer time satisfaction. The function graph is as follows: Figure 4 shown.

[0145] Assume that the service time of distribution center i to demand point j is t ij , then the time efficiency satisfaction function formula from distribution center i to demand point j is defined as follows:

[0146]

[0147] Product satisfaction:

[0148] Similarly, customers also attach great importance to the integrity of the product, and use the damage rate to measure the integrity of the product. Similar to the timeliness satisfaction, the freshness satisfaction function of fresh products at demand point j is constructed as follows:

[0149]

[0150] Then, the comprehensive customer satisfaction can be expressed as:

[0151] S j =γ1S j1 +γ2S j2 (2-8)

[0152] Among them, [L j ,U j ] represents the time that the jth customer expects to receive service; [δ1,δ2] represents the customer's satisfaction range with freshness; γ1,γ2 are weights. Through investigation, we know that cold chain enterprises pay more attention to the timeliness of the distribution process than to other costs.

[0153] In one embodiment of the present invention, the formula of the logistics distribution network optimization model constructed in this embodiment is:

[0154] MinZ1=min(F1+F2+F3+F4) (2-9)

[0155]

[0156]

[0157]

[0158] Where Z1 represents the total system cost of the logistics distribution network, F1, F2, F3 and F4 represent the leasing and operating cost of the logistics distribution center, the vehicle transportation cost, the customer delivery time penalty cost and the cargo damage cost respectively; S = {0,1,2,…,s} represents the set of distribution centers; C = {0,1,2,…,c} represents the set of ordinary goods customers; D = {0,1,2,…,d} represents the set of cold chain fresh food customers; M = {0,1,2,…,m} represents the set of ordinary goods distribution vehicles in the service area distribution center; N = {0,1,2,…,n} represents the set of cold chain distribution vehicles in the service area distribution center; l ij represents the delivery distance from point i to point j; u j represents the demand for ordinary goods of the jth customer; e j represents the cold chain goods demand of the jth customer; FS irepresents the daily rental of the distribution center in the i-th service area; H1 represents the handling fee per unit weight of ordinary goods in the distribution center; H2 represents the handling fee per unit weight of cold chain goods in the distribution center; FC k represents the fixed cost of dispatching the kth ordinary truck; FD k represents the fixed cost of dispatching the kth cold chain truck; f1 represents the energy consumption coefficient per unit distance of ordinary trucks; f2 represents the energy consumption coefficient per unit distance of cold chain trucks; f3 represents the penalty rate for providing logistics services before the customer's earliest time window; f4 represents the penalty rate for providing logistics services after the customer's latest time window; FRC ij It represents the highway toll required for ordinary vehicles to travel from point i to point j; FRD ij represents the highway toll required for cold chain vehicles to travel from point i to point j; X ijk , Y ij and Z i are all decision variables; [ET j ,LT j ] represents the time window for the jth customer to receive service; T jk represents the time when customer j receives the service from car k; rc j Indicates the damage rate of fresh products; rn j Indicates the damage rate of ordinary goods; p c represents the unit damage cost of cold chain fresh goods; p n It represents the unit damage cost of common goods;

[0159] X ijk , Y ij and Z i The value of is 0 or 1;

[0160] X ijk 1 means vehicle k has passed 0 before reaching point j, otherwise X ijk is 0;

[0161] Y ij If Y is 1, it means that the jth demand point is served by the i-th distribution center. Otherwise, ij is 0;

[0162] Z i 1 indicates that the i-th service area distribution center is selected, otherwise Z i is 0.

[0163] Formula (2-9) represents the total system cost of the logistics distribution network, including the leasing and operating costs of the logistics distribution center, vehicle transportation costs, customer delivery time penalty costs, and cargo damage costs.

[0164] Formula (2-10) represents the relevant costs of the logistics distribution center, including rental costs and operating costs.

[0165] Formula (2-11) represents the vehicle transportation cost, which specifically includes the fixed operating costs, fuel consumption costs, and highway tolls of different types of vehicles.

[0166] Formula (2-12) represents the delivery time penalty cost for ordinary goods customers and cold chain fresh food customers.

[0167] Formula (2-13) represents the cargo damage cost of ordinary goods and cold chain fresh goods.

[0168] In one embodiment of the present invention, the constraints of the constructed logistics distribution network optimization model are:

[0169]

[0170]

[0171]

[0172]

[0173]

[0174]

[0175]

[0176]

[0177]

[0178]

[0179]

[0180]

[0181]

[0182]

[0183]

[0184]

[0185]

[0186]

[0187]

[0188]

[0189] Among them, P k represents the maximum load of the kth ordinary delivery vehicle; Q k represents the maximum load of the kth cold chain delivery vehicle; r ijk represents the load of the kth vehicle when it leaves point i and arrives at point j; r jik represents the load of the kth vehicle when it leaves point j and arrives at point i; T ik represents the time when customer i receives the service of car k, t ik represents the length of time that customer i receives k-vehicle logistics service; t ijk represents the time required for the kth car to get from point i to point j; S j Indicates customer satisfaction, X jik is a decision variable, whose value is 0 or 1, X jik 1 means that vehicle k passed through point j before reaching point i, otherwise X jik =0, [ET i ,LT i ] represents the service time window of the distribution center in the i-th service area; T represents the minimum average comprehensive customer satisfaction.

[0190] Formulas (2-14) and (2-15) indicate that the load of each delivery vehicle at departure can meet the needs of all delivery customers on the route;

[0191] Equations (2-16) and (2-17) indicate that the vehicle capacity cannot exceed the maximum cargo capacity under any circumstances;

[0192] Equations (2-18) and (2-19) indicate that the load of all vehicles returning to the distribution center is 0;

[0193] Formulas (2-20) and (2-21) indicate that each customer is visited only once;

[0194] Equations (2-22) and (2-23) represent the change in the logistics service time that customer j receives from vehicle k;

[0195] Formulas (2-24) and (2-25) represent the order in which the vehicle passes through each node on the driving route, that is, passing through i first and then j;

[0196] Formulas (2-26) and (2-27) indicate that the vehicle departure time meets the earliest service time requirement of the distribution center;

[0197] Equations (2-28) and (2-29) represent the elimination of sub-loops in the circuit;

[0198] Equations (2-30) and (2-31) indicate that the vehicle must leave after serving the customer;

[0199] Formula (2-32) represents the average comprehensive customer satisfaction constraint;

[0200] Formula (2-33) indicates that the vehicle capacity is non-negative under any circumstances.

[0201] In one embodiment of the present invention, step S3 includes the following sub-steps:

[0202] S31, encoding the chromosome according to the number of vehicles, center numbers and customer point numbers in the logistics distribution network optimization model;

[0203] S32, initializing the population according to the encoded chromosome;

[0204] S33, calculating the comprehensive fitness value of chromosomes in the population;

[0205] S34, randomly performing genetic update on the initialized population by selection, crossover or mutation;

[0206] S35, using a local search algorithm to optimize the inherited population to obtain a new population;

[0207] S36. Repeat steps S33-S35 until a preset number of iterations is reached, thereby completing the iterative optimization of the logistics distribution network optimization model.

[0208] In one embodiment of the present invention, in step S31, the chromosome is encoded using a non-fixed-length real number encoding method.

[0209] Among them, a is the number of vehicles in use, a=1,2,3,...,m, m is the number of chromosomes; substring l w0 is the distribution center number, l w1 ,l w2 ,...,l wd Number the customer's order.

[0210] In this embodiment, there are six service area distribution centers as the distribution center candidate point set, twenty customer points need distribution services, and the number of vehicles currently in use is 5. First, the customer points are numbered: 1, 2, 3, ..., 19, 20, and the service area distribution centers are numbered: 21, 22, ..., 25, 26. A feasible chromosome is: V k ={22 2 1 3 5 6 23 4 7 1315 23 12 10 14 1124 19 9 8 24 17 16 20 18}, there are five substrings in this chromosome, namely: In the above five substrings, the service area distribution center numbers appearing are 22, 23, and 24, indicating that 22, 23, and 24 are determined as distribution centers in the distribution center candidate set; according to the arrangement of customer points, customer points 2, 1, 3, 5, and 6 are assigned to distribution center 22, customer points 4, 7, 13, 15, 12, 10, 14, and 11 are assigned to distribution center 23, and customer points 19, 9, 8, 17, 16, 20, and 18 are assigned to distribution center 24. In terms of the vehicle's driving route, the substring For example, the vehicle departs from distribution center No. 22, delivers to customer points No. 2, 1, 3, 5, and 6 in sequence, and finally returns to distribution center No. 22.

[0211] In one embodiment of the present invention, in step S32, the population is initialized in a random manner. First, one customer is selected from all customer points as a dividing point. Customers with numbers greater than or equal to the customer point number are arranged in the front in order, and customers with numbers less than the customer point number are arranged in the back in order to generate an initial sequence of customer points.

[0212] According to the load limit of each type of vehicle, the customer's needs are added to the route in sequence. When adding new customers, they are added according to the size of the left time window until the vehicle load exceeds the capacity limit and switches to the next vehicle.

[0213] In one embodiment of the present invention, the comprehensive fitness value calculated in step S33 can evaluate the ability of the current individual to adapt to the environment, and the calculated fitness value is used as the basis for whether to retain the chromosome to the next generation. The fitness function used in this embodiment is the inverse of the objective function F, that is,

[0214] In one embodiment of the present invention, the selection genetic algorithm used in step S34 is a roulette wheel method, which divides the roulette wheel according to the fitness value of the chromosome. The chromosome with a larger area on the roulette wheel is more likely to be selected and copied to the next generation.

[0215] In one embodiment of the present invention, in step S34, the method of genetically updating chromosomes in the population using a crossover genetic algorithm comprises the following sub-steps:

[0216] S3401, select two chromosomes in the population as parent chromosomes;

[0217] S3402, randomly select a coding position representing a distribution center in each parent chromosome, and extract the entire substring where it is located as the crossover position;

[0218] S3403, exchange and insert the crossover site into another parent chromosome to form a new chromosome;

[0219] S3404. De-duplicate the new chromosomes to obtain offspring chromosomes, thus completing the genetic update.

[0220] In this embodiment, there are six service area distribution centers as a set of candidate distribution center points, twenty customer points require distribution services, and the number of vehicles currently in use is five. The customer points are numbered: 1, 2, 3, ..., 19, 20, and the service area distribution centers are numbered: 21, 22, ..., 25, 26. Suppose a pair of parent chromosomes is: V1 = {22 2 1 3 5 6 23 4 7 13 1523 12 10 14 1124 19 9 8 24 17 16 20 18}, V2 = {22 7 4 2 8 21 6 5 3 1 21 12 10 916 26 13 15 17 20 26 19 14 11 18}, and it is assumed that {23 12 10 14 11} in V1 and {26 13 15 17 20} in V2 are randomly selected for crossover pairing. The crossover process is as follows: Figure 5 shown.

[0221] In one embodiment of the present invention, in step S34, the method of genetically updating chromosomes in the population using a variant genetic algorithm comprises the following sub-steps:

[0222] S3411, setting the mutation probability, and determining the chromosome to be mutated according to the mutation probability;

[0223] S3412, randomly selecting a number of substrings in the chromosome to be mutated;

[0224] S3413, performing distribution center code bit mutation on each substring;

[0225] The method for mutating the distribution center coding position is:

[0226] Randomly select a distribution center number that is greater than the number of customer points delivered in the substring from the candidate point set of the distribution center, and replace the distribution center number of the substring;

[0227] A344. Perform path encoding bit mutation on each substring to obtain offspring chromosomes and complete genetic update.

[0228] The method for performing path encoding bit mutation is:

[0229] A customer point number is randomly selected from the customer point numbers of the substring, and is exchanged with any customer point number of any substring in the same chromosome. When exchanging, it is necessary to ensure that the path encoding bits in the entire chromosome are not repeated.

[0230] In this embodiment, for

[0231] V1'={22 2 1 3 5 6 23 4 7 23 12 10 14 11 24 19 9 8 24 16 18 26 13 1517 20} chromosome, select the substring Perform distribution center code bit and path code bit mutation to obtain the mutated substring The mutated chromosome is: V1" = {25 2 1 3 5 8 23 4 7 26 12 10 9 11 24 1914 6 24 16 18 26 13 15 1720}. The mutation process is as follows: Figure 6 shown.

[0232] In one embodiment of the present invention, in step S35, the local search algorithm adopted follows the idea of ​​"destruction" and then "repair". First, a part of the customer elements in the current individual are removed by using the destruction operator, and then the repair operator is used to reinsert the previously removed customers into the individual, and the objective function is recalculated. The customers are inserted back to the position that makes the objective function as small as possible, and it is checked whether the current individual meets the vehicle load constraint and the time window constraint. If so, the current individual is updated to the new population.

[0233] In one embodiment of the present invention, demand points and distribution center candidate points of different sizes are randomly selected, and each algorithm is run ten times to obtain the average time, and the value of the optimal solution is recorded. The results are shown in Table 1:

[0234] Table 1 Algorithm performance comparison experiment

[0235]

[0236] The experimental results show that for Example 1, all three solving methods can obtain the optimal solution, and the running time of CPLEX is very short. The running time of the genetic algorithm combined with local search is slightly higher than that of the basic genetic algorithm.

[0237] For Example 2, the differences between each solution method begin to emerge. CPLEX, as a representative of the exact algorithm, can obtain the optimal total cost, and its result is better than the two genetic algorithms. However, due to the increase in the problem scale, its running time varies greatly and has exceeded the two genetic algorithms by a large margin. Since the two genetic algorithms use the number of iterations as the termination condition of the algorithm, their running time does not increase much. In terms of solving the optimal total cost, the genetic algorithm combined with local search is better than the basic genetic algorithm, but it takes a little longer.

[0238] For Example 3, when the number of demand points reaches 40, CPLEX can no longer find the optimal solution, while the genetic algorithm combined with local search can still obtain the result in a relatively short time, and the optimal value is better than the optimal value of the basic genetic algorithm. Through the experimental examples, it can be seen that the improved genetic algorithm combined with local search proposed in this method can better deal with the LRP problem in terms of solution time and solution quality.

[0239] In one embodiment of the present invention, taking the Biaoguo factory and a courier company as an example, it is assumed that the two companies intend to rent a distribution center in a joint distribution manner, and use the idle land in the highway service area to layout the logistics network to provide fast and high-quality urban logistics distribution services to community residents in the central urban area of ​​Chengdu, Sichuan Province ("12+2"). The goods it distributes include express parcels at room temperature and fruits that require cold chain distribution. The sources of the goods are all from outside Chengdu and have been transported by trunk lines to logistics nodes in Chengdu. Two types of vehicles, ordinary new energy trucks and cold chain new energy trucks, are used for urban distribution. Based on the above scenario, an analysis of the site selection and path problem of urban logistics distribution considering highway service areas is conducted.

[0240] The algorithm is implemented by Matlab R2016a programming, and the running platform of all programs is: Intel(R) Core(TM) i5-8265U CPU@1.60GHz 1.80GHz, memory 8GB, operating system windows 10x64.

[0241] (1) Site selection results

[0242] Results of selection of general distribution centers: 4 out of 12 candidate service areas were selected to expand general cargo distribution functions, including the West Ring Road Service Area, East Ring Road Service Area, Xindu Service Area, and Sancha Lake Service Area. The three distribution centers are located in the east, west, and north of Chengdu, and are all close to the city center, highlighting the geographical advantages of urban distribution. If a large distribution center is built in the corresponding location in the city, it will be costly and waste land resources.

[0243] Selection results of cold chain distribution center: 5 service areas were selected from 12 candidate service areas to expand cold chain distribution functions, including the West Ring Road Service Area, East Ring Road Service Area, Xindu Service Area, Sancha Lake Service Area, and Xinjin Service Area. The site selection results of the first four distribution centers are the same as those of ordinary distribution centers. These service areas do have superior geographical locations and can be built into comprehensive distribution centers. The site selection results are shown in Table 2:

[0244] Table 2 Service area distribution center selection results

[0245]

[0246]

[0247] Judging from the site selection results, this example requires the expansion of logistics functions in a total of 5 service areas in Chengdu. Among them, 4 will be built into comprehensive distribution centers, that is, providing both express parcel delivery and cold chain fresh food delivery services, namely the West Ring Road Service Area, Xindu Qingliu Service Area, East Ring Road Service Area, and Sancha Lake Service Area; 1 will be built into a cold chain distribution center, which only provides cold chain fresh food delivery services, and that is the Xinjin Service Area.

[0248] (1) The vehicle path results are as follows Figure 7 and Figure 8 The results of the express parcel delivery solution and the cold chain fresh food delivery solution are shown in Table 3 and Table 4;

[0249] Table 3 Results of express parcel delivery solution

[0250]

[0251]

[0252] Table 4 Results of cold chain fresh food distribution plan

[0253]

[0254] (3) Total system cost

[0255] In terms of total system cost, the cost of delivering express parcels is 15,411 yuan, and the cost of delivering cold chain fresh goods is 31,002 yuan, with a total cost of 46,413 yuan. In terms of customer satisfaction, the comprehensive customer satisfaction of express parcels is 91.2%, the comprehensive customer satisfaction of cold chain fresh goods is 90.4%, and the average comprehensive customer satisfaction is 90.8%.

[0256] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for optimizing a logistics distribution network, characterized in that: The following steps are involved: S1. Determine the set of candidate points for the city's distribution center; S2. Construct a logistics distribution network optimization model based on the distribution center; The formula of the logistics distribution network optimization model is: in, Z 1 represents the total system cost of the logistics distribution network, F 1. F 2. F 3 and F 4 respectively represent the leasing and operating costs of the logistics distribution center, the vehicle transportation costs, the customer delivery time penalty costs and the cargo damage costs; represents a collection of distribution centers; represents the set of general goods customers; Represents the collection of cold chain fresh food customers; It represents the collection of general cargo delivery vehicles in the service area distribution center; Represents the collection of cold chain delivery vehicles in the distribution center of the service area; Indicates from Click to Delivery distance of the point; Indicates The quantity of common goods demanded by each customer; Indicates The cold chain goods demand of each customer; Indicates Daily rental of a distribution center in a service area; It indicates the handling fee per unit weight of ordinary goods in the distribution center; It indicates the handling fee of cold chain goods per unit weight in the distribution center; Indicates The fixed cost of dispatching an ordinary truck; Indicates Fixed cost of dispatching a cold chain truck; It represents the energy consumption coefficient per unit distance traveled by ordinary trucks; It represents the energy consumption coefficient per unit distance traveled by cold chain trucks; It represents the penalty rate for providing logistics services before the customer's earliest time window; It represents the penalty rate for providing logistics services after the customer's latest time window; Indicates that ordinary vehicles from Click to Highway tolls required for point-to-point travel; Indicates that cold chain vehicles are from Click to Highway tolls required for point-to-point travel; , and All are decision variables; Indicates The time window in which a customer receives service; Indicates customer In accepting The time of the car service; Indicates the damage rate of fresh products; Indicates the damage rate of ordinary goods; Indicates the unit damage cost of cold chain fresh goods; It represents the unit damage cost of common goods; The constraints of the logistics distribution network optimization model are: in, Indicates The maximum load capacity of a common delivery vehicle; Indicates Maximum load capacity of a cold chain delivery vehicle; Indicates Car from Click to Load at the time of point; Indicates Car from Click to Load at the time of point; Indicates customer accept The time of car service, Indicates customer accept Length of vehicle logistics service time; Indicates Car from Click to The time required for the point; Indicates customer satisfaction, is the decision variable, Indicates The service time window of the distribution center in each service area; Indicates the lowest average comprehensive customer satisfaction; S3, use improved genetic algorithm to iteratively optimize the logistics distribution network optimization model; S4. Use the iteratively optimized logistics distribution network optimization model to layout the logistics distribution network and complete the logistics distribution network optimization.

2. The method for optimizing the logistics distribution network according to claim 1, characterized in that: The step S1 comprises the following sub-steps: S11, obtaining data of urban communities and determining original community demand points; S12, clustering the original community demand points using the K-medoids clustering algorithm to obtain a new demand point group; S13, clustering the new demand point group using the K-medoids clustering algorithm to obtain a number of cluster labels and cluster points; S14, select a cluster label, calculate the distance between all the points in the cluster label and each service area, select two areas with the smallest average distance to the points in the cluster as candidate points for serving the demand points in the cluster, and add them to the candidate list; S15, repeat step S14 until all cluster labels are traversed; S16. De-duplicate the candidate list to obtain a final set of candidate points, which is used as the set of candidate points for the city's distribution center.

3. The method for optimizing the logistics distribution network according to claim 1, characterized in that: The basic conditions for constructing the logistics distribution network optimization model include: The quantity and location of the distribution centers are determined from a set of candidate points; The location and demand of each customer demand point served by the distribution center are known quantities; The delivery vehicle departs from the distribution center to the customer demand point for delivery, and returns to the distribution center after serving all customers; Each customer demand point can only be delivered by one distribution center; There is no transfer of goods between distribution centers; The delivery vehicles travel at a constant speed on both highway sections and urban roads; Customers have time requirements for logistics delivery services, and exceeding the given time frame will incur time penalty costs; Each delivery vehicle corresponds to a delivery route, and the cargo load does not exceed the load limit.

4. The method for optimizing the logistics distribution network according to claim 1, characterized in that: The step S3 comprises the following sub-steps: S31, encoding the chromosome according to the number of vehicles, center numbers and customer point numbers in the logistics distribution network optimization model; S32, initializing the population according to the encoded chromosome; S33, calculating the comprehensive fitness value of chromosomes in the population; S34, randomly performing genetic update on the initialized population by selection, crossover or mutation; S35, using a local search algorithm to optimize the inherited population to obtain a new population; S36. Repeat steps S33-S35 until a preset number of iterations is reached, thereby completing the iterative optimization of the logistics distribution network optimization model.

5. The method for optimizing the logistics distribution network according to claim 4, characterized in that: In step S31, the chromosome is encoded using a non-fixed-length real number encoding method. ; in, a is the number of vehicles used, , is the chromosome number; substring , The distribution center number, Number the customer's order.

6. The method for optimizing the logistics distribution network according to claim 5, characterized in that: In step S34, the genetic algorithm used is the roulette wheel method, which performs roulette wheel division according to the comprehensive fitness values ​​of chromosomes in the population.

7. The method for optimizing the logistics distribution network according to claim 5, characterized in that: In step S34, the method of using the crossover genetic algorithm to genetically update chromosomes in the population includes the following sub-steps: S3401, select two chromosomes in the population as parent chromosomes; S3402, randomly select a coding position representing a distribution center in each parent chromosome, and extract the entire substring where it is located as the crossover position; S3403, exchange and insert the crossover site into another parent chromosome to form a new chromosome; S3404. De-duplicate the new chromosomes to obtain offspring chromosomes, thus completing the genetic update.

8. The method for optimizing the logistics distribution network according to claim 5, characterized in that: In step S34, the method of using a variant genetic algorithm to genetically update chromosomes in a population includes the following steps: S3411, setting the mutation probability, and determining the chromosome to be mutated according to the mutation probability; S3412, randomly selecting a number of substrings in the chromosome to be mutated; S3413, performing distribution center code bit mutation on each substring; S3414. Perform path encoding bit mutation on each substring to obtain offspring chromosomes and complete genetic update.

Images